AI’s Power Hunger Is Driving a New Wave of Energy IPOs

Lead

The AI trade is moving down the stack. After investors first crowded into chipmakers and large technology platforms, a new question is drawing capital: who will provide the electricity needed to run AI data centers?

According to Dealogic, energy companies raised $12.6 billion from initial public offerings in the first half of the year. That is the fastest pace this century and the highest half-year level since the peak of the dotcom bubble in late 1999. It also easily exceeds the $4.3 billion raised by the sector across all of 2025.

Key points

• Energy IPOs are surging: First-half fundraising reached $12.6 billion, a record first-half figure in Dealogic’s data.
• AI data centers are power-intensive: A typical AI-focused data center uses about 876,000 megawatt hours per year, roughly comparable to the household electricity consumption of Glasgow or Salt Lake City.
• Investors are shifting to “picks and shovels”: The market is looking for companies that supply generation, grid equipment, electrification infrastructure and distribution hardware.
• Public markets are funding heavier and riskier projects: IPO candidates include firms in electrical equipment, gas engines, nuclear power and next-generation geothermal.
• Performance is already split: Nearly two-thirds of energy companies that listed this year and last are trading below their IPO prices, according to Dealogic.

Why power has become an AI investment theme

AI infrastructure is not only constrained by access to GPUs. Large data centers also require grid connections, transformers, switchgear, generation capacity and sometimes on-site power. Consultancy ICF projects that U.S. electricity demand will rise 39 percent between 2026 and 2035, with data centers playing a major role in that increase.

That shift is reshaping investor appetite. Forgent Power Solutions, which makes electrical distribution equipment used in data centers, raised $1.7 billion in its February IPO. Innio, a German gas engine manufacturer, completed a nearly $2.8 billion listing in June as some data centers look to bypass strained grids and generate power on-site.

The enthusiasm is also reaching more speculative energy technologies. Fervo, which is developing next-generation geothermal using oil and gas drilling techniques, raised nearly $2.2 billion in May. The company plans to spend $1.2 billion over the next year on a Utah power station, and its leadership sees public markets as a way to accelerate growth.

Valuation also helps explain the rotation. Bloomberg data cited in the source shows the energy sector trading at about 18 times earnings, compared with roughly 40 times for information technology. For investors worried about whether hyperscalers can turn massive AI spending into profit, energy infrastructure offers a cheaper and more tangible way to play the same trend.

Significance and risks

The wave of energy IPOs suggests that the AI boom is entering an infrastructure phase. Power availability, grid bottlenecks, equipment lead times and local generation may become as strategically important as chips and models.

But the market is not treating all energy stories equally. Some companies are developing technologies that critics say are not yet technically or commercially proven. X-energy, ERock, Fermi and Deep Fission have all seen steep declines after listing. Deep Fission also raised far less than initially targeted.

The lesson is that “AI power” is a credible long-term theme, but not a guarantee of returns. Investors may increasingly separate companies with real customers, revenue and deployable products from those that remain closer to science experiments. The next phase of the trade may be less about buying every infrastructure IPO and more about identifying which businesses can actually deliver electricity at scale.

Source: Ars Technica AI